EP3552293A1

EP3552293A1 - Power control of inverters of a photovoltaic facility for participating in frequency regulation of the electrical distribution network

Info

Publication number: EP3552293A1
Application number: EP17811499.7A
Authority: EP
Inventors: Antoine ROSSÉ; Gauthier DELILLE
Original assignee: Electricite de France SA
Current assignee: Electricite de France SA
Priority date: 2016-12-12
Filing date: 2017-11-24
Publication date: 2019-10-16
Anticipated expiration: 2037-11-24
Also published as: FR3060229A1; CN110291694A; WO2018108481A1; US20190305561A1; CN110291694B; EP3552293B1; FR3060229B1; US10790673B2

Abstract

The invention concerns the control of electrical production by a facility of photovoltaic panels in order to constitute a power reserve, facility in which N inverters manage the production. A method is iteratively performed including the steps of: - transmitting (a) respective instructions to a number k of inverters among the assembly of N inverters, in order to each produce a maximum power, where k < N, the k inverters being determined for a given iteration of the method, and obtaining, from k inverters, measurements of maximum power produced respectively by the k inverters and recording (c) these measurements in a memory with N-k other measurements obtained from N-k complementary inverters during preceding iterations of the method; - estimating (d), by reading the contents of the memory, a maximum total power able to be produced by the assembly of N inverters of the facility, and obtaining a piece of information relating to a quantity of reserve power to be constituted, and determining (e) an overall power to be produced such that the difference between said overall power and said maximum total power corresponds to said quantity of reserve power to be constituted, - maintaining (f) the instructions of the k inverters to produce a maximum power, and transmitting respective instructions to N-k complementary inverters in order to produce a target power that is less than a maximum power, in order that the sum of the N-k target powers and the k maximum powers corresponds to said overall power to be produced, and - selecting, among the N inverters, one or more subsequent inverters for replacing the k inverters in a subsequent iteration of the method.

Description

Inverter power control of a photovoltaic plant for participation in the frequency adjustment of the electricity distribution network

The present invention relates to the management of a photovoltaic installation, of electricity production, comprising a plurality of inverters ensuring the connection between photovoltaic panels of the installation and an electricity distribution network.

It relates in particular to a method allowing such facilities to participate in the regulation of the frequency of the network, without recourse to a means of storing electrical energy, and without modifying the internal control of inverters currently available commercially. Currently, in France at least, the services needed to adjust the frequency (nominal frequency of 50 Hz in France) are mainly performed by conventional producers (in charge of production such as hydraulics, nuclear power, etc. .). Energy producers of the type commonly called "renewable energies" (wind, photovoltaic, etc.) are exempted. These frequency adjustment services are intended to help stabilize the frequency of the network around a nominal frequency (50 Hz in France). To do this, some producers (including conventional producers in France) are asked to:

- to increase the power in case of observation of under-frequency of the network (symptomatic of a lack of production or overconsumption in the system), - to decrease the power in case of observation of over-frequency (symptomatic over-production or lack of consumption).

A conventional producer can do this by setting the nominal setpoint of his installation below the maximum power of the plant. For example, if the frequency is below its nominal value of 50 Hz, for example 49.8 Hz, then it is desired that the installation produce more electrical energy (more precisely active power) to compensate the lack of production and to ensure that the frequency of the network returns to its nominal value of 50 Hz. Conversely, producers may be asked to temporarily reduce their production in case of observation of an over-frequency on the network (frequency above the nominal frequency, for example 50.2 Hz instead of 50 Hz in France, symptomatic overproduction of electrical energy in the system).

In addition, the dynamic frequency support, by a very fast release of the reserve, makes it possible to improve the quality of supply in the sensitive networks, like the island systems, by intervening very quickly following a disturbance. Today, in France at least, energy producers such as "renewable energies" (wind, photovoltaic, etc.) are exempted from participating in frequency adjustment.

Thus, photovoltaic installations are usually controlled so that the photovoltaic panels extract at every instant the maximum solar power available. This control is done by the internal control of each inverter, with algorithms that make it possible to find the optimal operating point of the solar panel (current, voltage) in order to extract a maximum of power. The type of algorithm usually used to achieve this control is called "Maximum Power Point Tracking" ("MPPT" below) and consists of finding the maximum power provided by each panel, depending on the voltage applied to its terminals. . The associated operating point is called "Maximum Power Point" ("MPP" below).

However, with the massive development of renewable energies on the electricity grids, new installations (including those with a "renewable energy" type) are increasingly being asked to participate in dynamic support and / or frequency tuning. .

In the case of an over-frequency, solutions exist to temporarily reduce the power produced by renewable energy producers. These solutions allow only to participate in the reserve "downward" (that is to say, to decrease production when the frequency increases). In the case of producers using wind energy, for example, it is possible to control the mechanical speed of the turbines or the pitch angle of the blades of wind turbines in order to reduce the power produced. In the case of photovoltaic installations, it is possible to reduce the power produced by moving away from the voltage corresponding to the operating point MPP of the panels supplying it. Such an embodiment is proposed in document FR 59128 in the name of the applicant. However, this strategy involves the internal control of the inverter connected to a panel or group of photovoltaic panels. However, the inverter is generally locked by the manufacturer so that it can only be used as a "black box" by the operator.

On the other hand, many commercial inverters remain controllable in power, that is to say that the operator can send an active power setpoint P _cons to the inverter, and this power will be produced at the output of the inverter (with a certain accuracy and a certain delay), if however this power remains lower than that P _M pp corresponding to its MPP. If the power setpoint P _cons is greater than P _M pp of the inverter, then the power produced will be limited to the maximum available power, that is to say the power P _M pp (called the "MPP" by the after). Specifically, an operator of a PV inverter can reduce the power produced by sending a lower power setpoint to the MPP of the inverter. Thus, this power P _cons can be considered as an active power limitation setpoint. In order to be able to support the network in the event of a drop in frequency, it is necessary to maintain an "upward" reserve, and thus keep a margin between the power available and the power actually injected into the network. The variability of primary resources (wind, sun, etc.) makes this reserve constitution particularly difficult. The technical solution often adopted to answer this difficulty and to contribute to dynamic support and / or frequency adjustment is to add a storage device within the installation. Nevertheless, this solution has as a major disadvantage an additional investment to be made in such a storage device.

The present invention improves this situation.

For this purpose, it is proposed an intrinsic modification of the operation of photovoltaic inverters and thus a modification of the usual modes of their control.

According to a first aspect, the invention proposes a method, implemented by computer means, for controlling the electrical power produced by a photovoltaic panel installation to constitute a power reserve, the installation comprising a set of N inverters for driving the power produced by said photovoltaic panels, the process, conducted iteratively on successive inverters among all the N inverters, comprising:

- Set respective orders to a number k of inverters among all N inverters, each to produce a maximum power, the number k being less than the number N and k inverters being determined for a given iteration of the process,

- Obtaining inverters from the maximum power measurements produced respectively by the inverters and store in memory said measurements for the inverters, said memory further storing Nk measurements obtained from Nk complementary inverters during at least one previous iteration of the process, - Estimate, by reading the contents of the memory, a total maximum power likely to be produced by all the N inverters of the installation,

- Obtaining information relating to a quantity of reserve power to be constituted, and determining an overall power to be produced so that the difference between said overall power and said total maximum power corresponds to said amount of reserve power to be constituted, - Maintain the instructions of the inverters to produce maximum power, and issue instructions respectively to Nk complementary inverters to produce a target power less than a maximum power, so that the sum of the Nk target powers and the maximum powers k corresponds to said overall power to be produced,

- Choose from the N inverters one or more following inverters to replace the inverters in a next iteration of the process. Thus, the present invention simply uses the current ability of known inverters to determine their operating point for delivering a maximum power, which thus makes it possible to determine the reserve in possible power of the installation, by cyclic determinations of the maximum powers of all the inverters. Some of these inverters (k inverters) are then driven in order to reach this maximum power while the other inverters (N-k inverters) can participate in the constitution reserve, and this successively (for example by circular permutation). The invention therefore only requires a computer device communicating with the inverters to control their operation according to such a method.

In an advantageous embodiment, a delay is applied between the emission of the setpoints to the inverters and the obtaining of the maximum power measurements of the inverters, in order to take account of a search latency by each inverter of the maximum power that can be produced. This research can be typically a pre-existing routine (called "MPPT") and described in more detail later.

In one embodiment, each instruction to the inverters to reach the maximum power comprises a time ramp setpoint to be respected, for a progressive increase in the power produced by each of the k inverters. Such an embodiment makes it possible to avoid peaks of power produced, beyond a global power setpoint to be respected, as will be seen below.

In one embodiment, before issuing a target power setpoint to a given inverter, it is verified whether said given inverter has received a maximum power production setpoint at the previous iteration of the method, and if necessary the setpoint target power to this given inverter comprises a time ramp setpoint to be respected for a progressive reduction of the power produced by this given inverter up to the target power. Here again, such an embodiment makes it possible to avoid unwanted reductions in power output, below an overall power requirement to be complied with. In this embodiment, combined with the previous one, the ramps of increase and decrease are chosen advantageously to keep on all the N inverters a global power produced corresponding to said global power to produce. This precaution makes it possible to smooth the overall power produced by making it conform to the calculated overall target, as will be seen in particular in FIG. 4B, commented on later.

In one embodiment, each target power setpoint is determined to respect the same difference between the target power and the maximum power for all the complementary N-k inverters. Such an embodiment makes it possible to distribute the reserve to constitute on the N-k inverters which are not at their maximum power produced.

In one embodiment, the aforesaid power reserve of the installation is constituted in order to free all or part of said reserve according to a frequency deviation with respect to a nominal frequency of an electricity distribution network to which the inverters are connected. In particular, the aforementioned information relating to a quantity of reserve power to be constituted may be a directly measured value of the frequency of the distribution network, according to which is calculated the amount of reserve power to constitute.

As a variant, this information relating to a quantity of reserve power to be constituted may be a reserve constitution instruction received from a managing entity of the distribution network.

In one embodiment, in a next iteration of the method, one of the N inverters is selected from the following inverters to replace previous inverters of a previous iteration (thus respecting the same number k from one iteration to the next). Typically, it is possible to provide a cyclic permutation of k inverters at each iteration, among the N inverters.

The present invention is also directed to a computer program (and its memory medium) comprising instructions for implementing the above method, when this program is executed by a processor. Figures 3 and 5 are possible flowcharts of the general algorithm of such a program. Figures 6 and 7 are possible flowcharts of particular subroutines of this general algorithm. The present invention also relates to a computing device comprising a processing circuit comprising a processor for implementing the method above. As illustrated in Figure 2 commented in detail later, the processing circuit further includes at least:

- a communication interface (ΓΝΤ) with the inverters, and - a memory (STO) for recording the maximum power measurements produced successively by the inverters.

Thus, the method enables the management of the active power setpoints of each inverter of the system, as a function of: the active power measurements of each inverter, practically in real time, and

- a real-time frequency measurement of the distribution network at the connection point (and / or a control signal from a network management entity).

This process allows pre-existing photovoltaic installations to be adapted, without replacement of current inverters or prediction of new sensors, to participate in the adjustment of the frequency.

Other advantages and characteristics of the invention will appear on reading the following detailed description of embodiments of the invention, and on examining the appended drawings in which: FIG. photovoltaic panels controlled by inverters, communicating with a device within the meaning of the invention,

FIG. 2 diagrammatically illustrates a device within the meaning of the invention,

FIG. 3 illustrates the main steps of a method in the sense of the invention, in an exemplary embodiment, FIG. 4A illustrates a frequency variation of the electricity distribution network, at the parameter input for a device of FIG. management of a three-inverter installation whose delivered powers are illustrated in FIG. 4B (bottom of FIG. 4B for the delivered individual powers),

FIG. 5 illustrates a detailed flow chart of an exemplary embodiment of a method according to the invention, FIG. 6 illustrates a detail of step S3 of FIG. 5,

FIG. 7 illustrates an embodiment detail of step S 12 of FIG. 5.

The present invention relates to the management of a photovoltaic panel installation as illustrated in FIG. 1. The installation comprises several photovoltaic panels connected by groups PV1, PV2, PVN to respective inverters OND1, OND2, ONDN. The panels produce electricity by photovoltaic effect and each inverter to which they are connected can be driven in power produced Ppl, Pp2, PpN at the output of each inverter, by a computer device (computer type) DIS. For this purpose, the device DIS can emit a setpoint CONS 1, CONS 2, CONSN, for each inverter OND 1, OND2, ONDN. In particular, this setpoint for an inverter can be to reach a maximum power produced PMPP- Once this power value P _M pp reached by an inverter, the latter usually transmits to the DIS device the maximum power reached PMPP- Thus, the DIS device can know at any time a measure of this maximum achievable power PMPP- This value P _M pp is variable and of course depends on the conditions of sunshine (it is high as the sun is important). Thus, the value P _M pp that each inverter transmits to the device DIS represents a measurement (arrows MES 1, MES 2, MESN) at a current time of the maximum power that can be reached and that can be delivered by this inverter (and thus depending on the degree of sunshine at this current moment). More specifically, the inverters are usually controlled so that their photovoltaic panels extract at every moment the maximum of the available solar power. The type of control algorithm usually used to achieve this control is called "Maximum Power Point Tracking"("MPPT") and consists of finding the maximum power provided by each panel, depending on the voltage applied to its terminals. The associated operating point is called "Maximum Power Point"("MPP"). It is also possible to reduce ("clamp") the power produced by an inverter by moving away from the voltage corresponding to the operating point MPP of the panels supplying it. In this case, the power produced Pp is less than the power PMPP. In practice, the operating point control algorithms of the photovoltaic panels are integrated in a control unit of the inverters which are commercially available and are locked by the manufacturers. . Thus, for a user, the inverter behaves like a "black box", with a limited choice of external control modes. Nevertheless, most inverters offer at least two control modes: - MPP control: the inverter operates in MPP mode to deliver the maximum power P _M pp at a current time and uses an MPPT type algorithm to extract the maximum power at any time,

- Control by power Pcons command: the inverter receives a power instruction, for example from an external device, and tries to achieve this. If the available solar power is sufficient, the inverter then seeks an operating point of the photovoltaic panel making it possible to achieve the requested power setpoint (for example by modifying the voltage at the terminals of the panel with respect to the voltage giving the maximum power P _MPP ). It can be considered in this case that the inverter is "clamped" because it does not exploit the panel to the maximum and therefore does not extract the maximum power available. However, it does provide a power corresponding to the instruction sent to it. On the other hand, if the power setpoint sent to the inverter is greater than the available solar power, the inverter seeks to operate at the maximum available power, that is to say in "MPP" control as long as this maximum power available remains below the set point. In the context of the present invention, it is considered that the inverters can each receive:

a power setpoint for clamping the power of the panel if the available power allows it, and

- an order to operate at the MPP, and thus find the maximum power point through the MPPT type algorithms implemented in the internal controller of the inverter. The invention then proposes to take advantage of these possible modes of operation to constitute a reserve of electrical power generated by the panels and thus contribute to the balance of the frequency of the electrical distribution network (as the purpose described in document FR 15 59128 ). For example, it is necessary to correct a difference in frequency Af with respect to a nominal frequency (in France of 50 Hz). In principle, the higher the frequency (Δ /> 0), the more it is necessary to reduce the power produced. The lower the frequency (Δ / <0), the more it is necessary to release the power reserve produced. In an extreme case where Af <0, one can control the inverters so as to deliver the maximum power P _M pp of all the inverters.

For this purpose, the device obtains a current value of the frequency of the network, or a deviation at its nominal value Af. To obtain this datum, with reference to FIG. 2, the device DIS can receive (or alternatively directly measure) the value Af via a communication interface COM. The device further comprises a processor PROC and a permanent memory MEM. This memory MEM can typically store instruction codes of the computer program within the meaning of the invention. These instruction codes can be read by the processor PROC to perform the method according to the invention (described in detail below). The device may furthermore comprise a working memory STO for storing temporary data, in particular the maximum powers P _M pp that the inverters bring back to the device via the communication interface with the inverters ΓΝΤ, which can also be used to send the instructions. of power to be produced to the inverters. It should be noted that the memories MEM and STO can be a same memory unit of the device. In addition, particularly for an optional step (but advantageous timing) of the method presented below, the device may include an internal clock HOR.

In particular, it is sought to build up the power reserve on the photovoltaic installation and optimize its constitution, so as to limit the lost output and maximize the availability of the reserve. The invention thus aims at a method, implemented by computer means (typically the DIS device), of power generation control of photovoltaic panels by sending precisely calculated active power setpoints to the individual inverters of an installation (called " central "below), in order to have at all times an estimate of the maximum available power of the plant (called" global MPP "of the plant) and to be able to achieve a global power requirement, less than or equal to the overall MPP . This method does not involve any modification of the internal controllers of the inverters (for example an algorithm that shifts the MPP for example, as described in document FR 59128), which can be considered as "black box" type systems with a setpoint input power and output power realized and measured. The process can be compatible with existing installations using commercially available inverters, provided that the inverters can receive power instructions from an external device (eg a PLC, a computer, etc.), and that their time response to these instructions is sufficiently short (typically less than 5 seconds).

One embodiment of the method can be described by the steps illustrated in FIG. 3, implemented cyclically for an installation comprising a total number N of inverters: a) Send an order to a number k of inverters among the set N inverters (k greater than or equal to 1 but less than N), to produce their maximum power (setpoint to go to the MPP),

b) Apply a chosen delay (so that the search of the MPP by the internal search algorithm to these inverters has converged towards the MPP point and that the powers produced by these inverters are indeed the respective MPP powers), c) Measure the powers produced by the inverters at their maximum power and record these values (these power values indicate the MPP of these inverters at this time and are stored in a memory for use as estimates of the current MPPs of these inverters if they are later curbed), d) Calculate an estimate of a global maximum power (called global MPP) likely to be produced by the facility (by summing the recorded MPP estimates from the facility). set of N individual inverters),

e) Possibly in parallel with steps a) to d), obtain a data of the frequency of the electricity transmission network and calculate a global power setpoint to release all or part of the reserve of the installation according to a deviation of frequency relative to a nominal frequency (denoted hereinafter Af) (this frequency of the network and therefore its deviation, can be measured directly by the installation, or alternatively received in the form of a release order of all or part of the reserve, for example in a signal received by the installation, and from a network management entity for example), f) Send power instructions to the (N - k) inverters (inverters of the installation which are not at their MPP), to clamp down (almost instantaneously) their power, so that the total power of the installation follows the global power setpoint (less than or equal to the overall MPP of the installation): these inverters ensure nsi the active power adjustment or the adjustment effort of the installation at grid frequency, while the other k inverters are at their MPP,

g) Repeat the previous steps by modifying the inverters which are in the MPP and those which are clamped (by choosing the following inverters for example among the N inverters): the inverters can each go in turn to the MPP and update the value estimated MPP of each inverter cyclically or according to advanced alternative strategies, while maintaining the reserve on the rest of the inverters which are thus clamped power.

Step f) can be performed in different ways: the (Nk) inverters can all have the same individual power setpoint, or alternatively have different power setpoints from one inverter to another, but for the same purpose to produce a total power of the installation respecting the overall power setting. Typically, some of the inverter (N-k) inverters may for example be at their MPP, while others may be below.

Thus, a global reserve is established at the plant by placing at each instant a selected number Nk of inverters at a power less than or equal to their MPP, by sending them power orders less than or equal to the estimates of their respective MPPs. . The MPP estimates are cyclically updated to ensure rotation between inverters that are in the MPP and those that perform the reserve build function. Thus, an estimate of the overall MPP of the plant can be known at any time (taking all measured and recorded MPP values one after another on all N inverters). The overall power can then be controlled to provide reserve and adjust the frequency of the network, without requiring sensors other than the usual power sensors of the inverters (the inverters trade being already equipped with these sensors). All that is required is a sensor of the network frequency at the connection point of the control panel, or simply of a communication module for receiving an external adjustment signal (for example a signal coming from a network management entity). . No modification of the internal controllers of the inverters nor of a priori knowledge of the behavior of the panels is necessary.

As an illustrative example, in the simple case of an installation consisting of three inverters, each with a maximum power of 10 kW, the total power of the plant with full sun is therefore 30 kW. By participating in the frequency regulation contribution of the network, this maximum is never reached (except in the case of full sunlight and where the frequency is well below 50Hz). Rather, it is sought in this example to maintain a fixed power reserve of 4 kW, which could be released quickly in the event of a significant drop in the network frequency or activation by the setpoint of the aforementioned external signal. It is considered that at an initial state, the estimates of the MPPs of each inverter are known and equal to 6 kW for each inverter. The overall MPP estimate of the MPP _is _ _tota i is the sum of the estimated MPPs of the three inverters. So MPP _is _ _tota i = 18 kW.

During the first phase of the cycle, the OND1 inverter is controlled to reach its MPP. It is expected to produce 6 kW, and it seeks to maintain 4 kW of reserve on the rise. It is therefore necessary to produce a total of 18 kW - 4 kW = 14 kW, that is to say that the 4 kW reserve must be placed on the OND2 and OND3 inverters. By distributing the reserve equitably, in this example, 2 kW of reserve is placed on each inverter (OND2 and OND3), so each receives a setpoint Pcons = 4 kW for example. Other distribution solutions are possible, for example to optimize the efficiency of the inverters. The total measured power is 14 kW, given that the overall MPP estimate is 18 kW. Indeed, the power MPP of the inverter 1 has not changed compared to its estimate. So we have 4 kW of reserve in theory, as follows:

- OND1: MPP = 6 kW and Pcons = MPP = MES1 = 6kW,

- OND2: MPP = 6 kW, but Pcons = 4 kW = MES2, and - OND3: MPP = 6 kW, but Pcons = 4 kW = MES3, with MES1 + MES2 + MES3 = P, = 14kW, while MPPglobal = 18 kW, so 4 kW reserve for this plant.

By switching to the second phase, the OND2 UPS is now operating at the MPP (instead of the OND1 UPS), and the reserve is now placed on the OND1 and OND3 inverters. In the example described, there is a difference between the estimated MPP and the measured power of the OND2 inverter: its MPP has changed to 6.5kW. The new MPP value of the OND2 UPS is then recorded at the end of this phase to update the overall MPP and correct the next phase setpoints to maintain the 4 kW reserve. At the end of the second phase, we have:

- OND1: MPP = 6 kW and Pcons = 4kW,

- OND2: MPP = 6.5 kW and Pcons = MPP = 6.5kW, and

- OND3: MPP = 6 kW and Pcons = 4kW.

Thus, P _t otai = 14.5 kW and global MPP = 18.5 kW, and we have 4 kW reserve for this plant. During the third phase, the OND3 UPS goes to the MPP. It is found that its MPP is unchanged.

We have at the end of this third phase:

- OND1: MPP = 6 kW and Pcons = 4kW,

- OND2: MPP = 6.5 kW, so that Pcons can be chosen at Pcons = 4.5kW, and - OND3: MPP = 6 kW and Pcons = 6kW.

It has _to pillowcase P ⁼ 14.5 kW global MPP = 18.5 kW with well 4 kW subject always to the plant.

The UPS OND1 is then switched back to its MPP to restart the cycle, and update the MPP of the OND1 inverter to the typical case where it has evolved since its last measurement. It is noted with this cyclic algorithm MPP setting of successive inverters that it is possible to maintain a reserve volume on the plant while following the variations of the MPP of the individual inverters and distribute a global setpoint on all inverters. The accuracy of the reserve varies with changes in sunshine that result in changes to the MPP. It is therefore best to scan all the inverters cyclically and update all the MPPs regularly. Since variations are slow, precision The estimation of MPPs is good and a selected reserve can be guaranteed with satisfactory accuracy.

FIGS. 4A and 4B are shown to show a test of the process under real conditions on an installation composed of three inverters (each with a maximum power of 15 kW), driven in power by a computer DIS (automaton) which sends instructions of active power to the inverter according to the power measurements received from the inverters. A frequency profile of the network f, of synthesis, is used (FIG. 4A) to test the response of the installation (Ppl + Pp2 + Pp3 corresponding to the sum of the powers produced by all the three inverters) to variations of network frequency. FIG. 4B validates the correct multi-inverter operation according to the method of the invention.

The analysis of this test clearly illustrates that the inverters are successively controlled, one by one, at the MPP in order to find their maximum power point and update the estimated MPP value and kept in memory. On the curves of the individual power measurements of the inverters, there are ramps that the inverters follow to go up and down their MPP to make the power transitions softer. This embodiment of "ramp" variation is described in detail below. The noise at the level of the overall power is low and the overall power follows the global instruction with very little error at every moment. The MPP estimate is correct because at times when the frequency is very low (at t = 400s and then just before 600s in the example shown), the entire reserve is released (that is, the central is set to the MPP), and the total power obtained (Ppl + Pp2 + Pp3) is the one expected and corresponding to the overall Consglobal setpoint, with very little error except the maximum power requested since it can not exceed the sum P _M p of the three inverters.

We note the proper functioning of the frequency adjustment: the overall power responds well to variations in frequency. When the frequency increases, the power drops, and conversely when the frequency drops, the power of the power station increases to support the network. The curve (Ppl + Pp2 + Pp3) representing the variation of the overall power produced as a function of these frequency variations of the network follows well the set point and thus shows that it is possible to perform a quality adjustment.

More specifically, specific embodiments are described below as exemplary embodiments. The control of the power of the inverters is based on the principle of the cyclic process presented previously with reference to FIG. 3. The control device DIS keeps in memory a table or a vector of "Tokens", which indicates at each instant which inverters must go to the MPP and which ones should ensure the reserve. For example, if the first element of the vector Token (l) = 1, then the UPS OND1 is controlled at the MPP. If Token (l) = 0, then the OND1 inverter is power-clamped and receives a power setpoint to form part of the plant reserve. At the end of each phase of the cycle, the Tokens are "redistributed" to change the distribution between the inverters at the MPP and those providing the reserve. Those who provide the reserve receive a setpoint calculated precisely so that the global instruction is performed (function of the desired reserve, the maximum available power, and the frequency of the network). These calculations are detailed below. In addition, as introduced above, an additional feature, called "setpoint ramp smoothing" is added to smooth the power transitions between the "reserve" mode and the "MPP" mode of each inverter, while ensuring the realization of the global instruction at any time. This feature is also detailed later.

In general, we can note three cases in which all the inverters are forced to the MPP:

the first case concerns the initialization phase to recover all the MPPs of all the inverters,

the second case occurs when the estimated overall MPP is below a threshold predefined by the operator (insufficient power available to provide the reserve), and

the third case occurs when the global power setpoint is equal to or greater than the overall current MPP of the plant (thus seeking to release all the available power, as previously illustrated in FIG. 4B at times 90s, 400s, and 590s).

In this example, a cyclical scan of one of three inverters is presented. Other scanning strategies are possible to improve the estimation of the MPP or to have at any time the maximum controllable power. Some of these strategies can be based for example on the measurement of the DC voltage at the terminals of the panels. In addition, in the case of two inverters only for example (or respectively four inverters), one (respectively two) inverter can be placed at its (their) MPP, while the other (both). others) follows (follow) an imposed power setpoint. In addition, each inverter can have its own MPP, different from the MPP of another inverter (because of the number of panels to which it is connected, or their own degradation). FIG. 5 illustrates the algorithm corresponding to an embodiment detail integrating the first initialization step S1, presented above. For a duration T _init , the inverters are all controlled at the MPP. The time T _init must be long enough for the MPPs to stabilize (time required for the internal MPPT algorithm of each inverter to converge). This value depends on the inverter model and the manufacturer's specifications, but is in the order of 10-20 seconds for a UPS in commerce today. No reservations are ensured during this phase. At the beginning of the initialization (t = t ₀ ), the first value of the variable t _fm _ _{cyde is defined} . Thus, the initialization ends as soon as the initialization time has elapsed, and the frequency setting can actually begin in the following steps. Previously, the estimated values of the MPPs of each inverter are stored in the memory of the DIS control device.

These values are updated in step S3 as soon as each "phase" of the cycle is complete (OK arrow at the output of the test S2), and a number k of inverters have been ordered at the MPP to obtain a value update of their MPP. The block diagram for this step S3 for updating MPP estimates is presented in FIG. 6 commented below, for a number N of inverters. In step S31 of FIG. 6, an inverter counter i (i = 0) is initialized. In test S32, as long as the counter i is smaller than the total number of inverters N (arrow OK), it is possible to update the values of MPP stored with measurements obtained in the last cycle performed. If the counter i has reached the value N (arrow KO), then, in step S33, this routine for updating the MPP may stop. Otherwise, in step S34, the counter i is first incremented (to designate a next inverter), then in step S35, a test is made to determine if this inverter i is currently placed at its MPP, to which case (arrow OK-S35) the output power measured on this inverter Pmes (i) corresponds to its estimated MPP (MPPest (i)) and can be stored in memory in step S36. Otherwise (arrow KO-S35), the estimated MPP of the current inverter i remains unchanged, as it is currently stored in the memory of the DIS device (step S37).

Returning to FIG. 5, in step S4 (which follows the MPP update step S3), the cycle parameters are reset to prepare the next phase of the cycle. The Tokens are also redistributed to indicate the next inverters to be placed at their MPP. The previous value of the tokens is kept in memory to determine the inverters placed in the MPP at the previous phase of the cycle. This step S4 is thus carried out at:

a redistribution of the Tokens: the state of the tokens in the previous phase having been recorded and kept in memory, the chips are redistributed by circular permutation (for example, for a central with four inverters with two tokens and two inverters at the MPP at the times, we would have 1001 that becomes 1100 if bit 1 defines an inverter to be placed at the MPP); and - after the definition of the new Tokens, the new slopes are calculated for the smoothing ramps of the power transitions (calculation of the new ramp parameters). This calculation is detailed below.

In step S5, the time variables defining the end of phase of the ramps and the end of update phase of the MPP of the inverters are redefined, as follows:

O tfin ramp _ram T t p _e O tfin cycle t T _C y _C i _e

In particular, it is a matter of determining a moment when a power smoothing ramp can be completed and a moment when the update of the MPP can be considered complete.

Of course, the cyclical distribution of the Tokens, presented here, is only one embodiment that can be used for controlling the inverters at the MPP. There are other possible realizations to redistribute Tokens.

Moreover, as long as a cycle is not completed (KO arrow at the output of the test S2), the estimated cycle parameters and MPP remain unchanged (step S6).

In step S7, the overall MPP of the plant is estimated by summing the estimates of the individual MPPs of all the N inverters of the plant:

MPP _globa l_est = Σ MPP _is (l)

UPS

Then, in test S8, this _global MPP value is compared to a threshold THR activation of the reserve procedure according to the invention (regardless of the frequency balance of the network). If the _overall maximum MPP power is below this THR threshold (KO arrow at the output of the S8 test), then all the inverters are simply forced to operate at the MPP in step S9.

If not (OK arrow at the output of the test S8), a global power instruction P (f) which can be lower than the _global maximum MPP is calculated and depends on the frequency of the network f, in the step S10.

As an example of calculating P (f) for a primary adjustment of the frequency, the global setpoint is calculated from the total available power (estimation of the overall MPP), the desired "average" reserve quantity (Ro ) and the frequency measurement of the network, to make a primary adjustment of the frequency, as follows:

Pcons _total = _global MPP _is - R ₀ - H (f - f ₀ ) (with R ₀ and H greater than 0).

To this calculation, saturation is added to preferentially respect the limits below: MPP _is _ total 2R ₀ ≤ _total Pcons ≤ _global MPP is

Such an embodiment makes it possible to ensure that the power that can effectively be produced as much as possible by the central unit can not be lower than the given power setpoint (situation of FIG. 4B at 90s, 400s, etc.) and thus maintain coherence between the power setpoint and the power actually produced. Thus, setting the value of H defines the "normalized" power of the control unit for the frequency setting. The overall power setting can increase to respond effectively to a decrease in frequency and decrease to respond to an increase in frequency.

Of course, this is just a simple example for frequency tuning with proportional gain. Other embodiments for the frequency adjustment are possible, for example with variable slopes and / or the addition of dead band. The global setpoint can also be calculated as a function of an external signal coming for example from a network management entity requesting to release more or less reserve at each instant.

As long as the global power setpoint remains below the maximum possible output MPP _g iobai_est (arrow OK at the output of the test SU), the individual power instructions CONS (i) are calculated and sent to the inverters, as follows, at the step S 12, with here preferably a search for power distribution of the non-MPP setpoint. Otherwise (KO arrow at the output of the test SI 1), all the inverters are forced to operate at the MPP at the step SI 3.

In step S12 (detailed below with reference to FIG. 7), the individual powers are calculated so that the total power actually delivered is equal to the global power setpoint. They are also calculated so that the inverters that go from a clamped setpoint to the MPP can gradually increase to the MPP (by respecting a smoothing ramp for the power transitions), and that the inverters that "go down" from their MPP can also go down gradually to avoid power jumps. This embodiment smooths the power transitions. This feature is detailed below. To avoid sudden transitions of individual powers due to setpoint jumps (switching from a clamped power to the MPP and vice versa), a ramp is added to the power setpoints to "soften" these transitions. Indeed, it has been observed, in particular for some inverters, a power growth to reach the MPP which can be faster than a decrease in power since the MPP. It is possible for other models of inverters that this trend is reversed. In any case, this difference in speed of variation to reach the MPP generates peaks of overall power produced relative to the setpoint, which should then be attenuated. The effect of such ramps on the individual powers produced in the example of FIG. 4B can be observed (for the powers Ppl, Pp2 and Pp3). The goal of ramps is that power global is always respected but the power transitions at the individual inverters are "softer", respecting that the sum of the instructions always equals the global setpoint.

In an exemplary embodiment, the slope of each inverter is calculated as follows:

ΔΡ (ί)

P_rampe i) =

T ramp ΔΡ (ί) is the difference between the MPP of the inverter (i) and its power at the instant before the beginning of the ramp. In practice, it is calculated using the estimate of the MPP of the inverter (last value kept in memory) and the last power setpoint sent to the inverter. This calculation is done in step S4 presented previously with reference to FIG.

For a set of inverters (possibly with several inverters that go up to the MPP at the same time as several inverters descend from the MPP), each inverter that goes up to the MPP receives a ramp setpoint before reaching its MPP, and the inverters that go down receive a setpoint which decreases progressively from the MPP towards a restrained setpoint with a view to the reserve. At each moment the instructions are calculated in order to respect the overall power of the plant.

The total ramp setpoint is also defined: P_rampe_tot = Σ ₀ nduieurs P _ro pe (V)

amounts

At time t, if the inverter i is powering up to its MPP, its new setpoint is given by the relation:

(RI): Pcons (i) (t) = _MPP (i) - (t _{ramp end} - t - T _ech) * P_ra pe {i) where T _ec h is the sampling time. This formula is valid regardless of the sampling time of the algorithm. The power reserve is ensured by inverters that are not at the MPP. They each receive a setpoint, proportional to their power of MPP, calculated so that the sum of the powers produced by the inverters respects the global instruction.

For an inverter that is not at the MPP, its power setpoint is given by the following relation:

Pcons_globale -Σ _is duleurs MPP _is (i)

(R2): Pcons) = - y ^ΜΡΡ _{MP P m} * ^MPP is U)

^MP P is total-Londuleurs ^{Mp p} estl ^l )

at the MPP

The estimated values of the MPPs (kept in memory) are used to calculate the instructions. Thus, with the assumption that the power instructions are well realized, the sum of the individual powers gives the global power setpoint. Furthermore, in order to compensate for the power ramp of the inverters that go up to the MPP, the setpoint of the inverters that "go down" must be adapted, so that at each instant the sum of the individual instructions gives the global instruction. Thus, at time t, if the inverter i is going down from its MPP in order to respect the reserve, its new instruction is given by the relation:

(R3): Pcons (j) (t) =

^u is descending from the MPP where T _ec h is a given sampling time.

Using these relationships and summing up the instructions:

- inverters which are in the MPP, - inverters which participate in the realization of the reserve,

- inverters rising to the MPP, and

- Inverters coming down from the MPP, we find that the sum of the instructions is equal to the global instruction at time t.

FIG. 7 illustrates an algorithm flow diagram corresponding to this treatment imposing the aforementioned ramps. After a step S131 of initialization of an inverter counter i (i = 0), a comparison test S132 to the number N (with if i = N the end of this routine in step S133) and an incrementation of i in step S 134, it is first determined whether the token of the inverter i is 1 (this inverter must be brought to its MPP) in the test S 135. It is further determined whether the previous token JET-1 (i) of this inverter (at the previous cycle) was already at 1 in tests S136 and S137. It is furthermore determined whether the current time has exceeded a selected delay t _ramps to tests S138 and S139.

If the current token of inverter i is 1, and if the previous token of inverter i is equal to 1, then it is not necessary to make a power variation to be produced by this inverter i and to step S 140, its power setpoint Pcons (i) corresponds to its MPP.

If the current token of the inverter i is 1 but if in addition the previous token of the inverter i was 0, while the delay t _ramp has not yet elapsed, then in step S141, the driver is inverter i to respect the ramp up to its MPP with its power setpoint Pcons (i) calculated according to the relationship RI. In this case, however, after the timer t _{ramp is} complete, step S140 is applied by requiring the inverter i to reach its MPP. If the current token of the inverter i is 0 while the previous token of the inverter i was 1, then it is necessary to "down" the inverter i of its MPP and a delay is applied until t _ramp . During this delay, the power setpoint Pcons (i) of this inverter i is determined by the relation R3 given previously and which depends on the ramp imposed, in step S142. Then, once the delay is over, the power setpoint no longer depends on the imposed ramp and is determined by the relationship R2.

If the current chip of the inverter i is equal to 0 and if the previous token of the inverter i is equal to 0, then in the step S 143, its power setpoint Pcons (i) remains lower than the MPP and is determined by the relation R2 given previously.

By the implementation of the invention, a photovoltaic farm composed of several inverters can participate in the dynamic support of the frequency without adding storage or additional sensors, and without modifying the internal controllers of photovoltaic inverters currently available commercially. The algorithm for cyclic estimation of the MPPs of the inverters makes it possible to dynamically estimate the overall MPP of the plant, and to follow the variations of MPP, while producing an overall power lower than the MPP of the plant, and thus to maintain precisely a volume of reserve (especially if the dynamics of sunshine variation remain slow). Such a result is achieved without any knowledge model, sensor, or modification of the internal controllers of commercially available PV inverters. The knowledge of the MPP at any time allows the manager of the electricity distribution network to know the volume of spare available to meet the distribution need. A simple computer means (automaton or computer for executing the algorithm) comprising suitable communication buses (to communicate with the inverters, such as the interface ΓΝΤ of FIG. 2) allows the implementation of the invention on photovoltaic farms existing inverters already having the capacity to receive active power setpoints and to return their active power measurements. Thus, some existing photovoltaic plants (with inverters capable of receiving instructions from an external device) can be easily adapted to perform frequency adjustment without modification of the inverters or photovoltaic panels. The method of the invention can easily be adapted to operate with any number of inverters (greater than two), and to operate with inverters having different dynamic characteristics, different powers (with a simple adjustment of the parameters of the algorithm ). The algorithm can be used for the implementation of other services requiring a control of the photovoltaic production (other than the contribution to the balance of the frequency of the network).

Claims

1. Method, implemented by computer means, for controlling the electrical power produced by a photovoltaic panel installation to constitute a power reserve, the installation comprising a set of N inverters for controlling the power produced by said photovoltaic panels , the method, conducted iteratively on successive inverters among all the N inverters, comprising:

- Set respective orders to a number k of inverters among all N inverters, each to produce a maximum power (a), the number k being less than the number N and k inverters being determined for a given iteration of the process ,

- Obtaining the inverters of the maximum power measurements produced respectively by the inverters and recording in memory the said measurements for the inverters (c), the said memory also storing Nk measurements obtained from Nk complementary inverters during at least one previous iteration of the process, - estimate, by reading the contents of the memory, a total maximum power likely to be produced by all the N inverters of the installation (d),

- Obtain information relating to a quantity of reserve power to be constituted, and determine an overall power to be produced so that the difference between said overall power and said total maximum power corresponds to said amount of reserve power to be constituted (e) Maintain the instructions of the inverters to produce a maximum power, and issue setpoints respectively to Nk complementary inverters to produce a target power less than a maximum power, so that the sum of the Nk target powers and k maximum powers corresponds to said overall power to be produced (f),

- Choose from the N inverters one or more following inverters to replace the inverters in a next iteration of the process.

2. Method according to claim 1, wherein a delay (b) is applied between the transmission of the setpoints to the inverters and the obtaining of the maximum power measurements of the inverters, to take account of a search latency by each inverter of the maximum power that can be produced.

3. Method according to one of the preceding claims, wherein each instruction to k inverters to achieve the maximum power comprises a time ramp setpoint to be respected, for a gradual increase in the power produced by each k inverters.

4. Method according to one of the preceding claims, wherein, before the issuance of a target power setpoint to a given inverter, it is checked whether said given inverter has received a set of maximum power output at the iteration previous method, and if necessary the target power setpoint to this given inverter includes a time ramp setpoint to be respected for a progressive decrease in the power produced by this given inverter to the target power.

5. The method of claim 4, taken in combination with claim 3, wherein the ramps of increase and decrease are chosen to keep on all N inverters a global power produced corresponding to said overall power to be produced.

6. Method according to one of the preceding claims, wherein each target target power is determined to meet the same difference between the target power and the maximum power for all N-k complementary inverters.

7. Method according to one of the preceding claims, wherein said power reserve of the installation is constituted in order to release all or part of said reserve according to a frequency deviation (Δ /) with respect to a frequency nominal of an electricity distribution network to which the inverters are connected.

8. The method of claim 7, wherein said information relating to a quantity of reserve power to be constituted is a measured value of the frequency of the distribution network, according to which is calculated said amount of reserve power to constitute.

9. The method of claim 7, wherein said information relating to an amount of reserve power to be constituted is a reserve constitution instruction received from a manager entity of the distribution network.

10. Method according to one of the preceding claims, wherein, in a next iteration of the method, one of the N inverters, k following inverters are chosen to replace k previous inverters of a previous iteration, by performing a cyclic permutation of k inverters at each iteration.

11. Computer program characterized in that it comprises instructions for implementing the method according to one of claims 1 to 10, when the program is executed by a processor.

12. A computer device comprising a processing circuit comprising a processor for implementing the method according to one of claims 1 to 10, the processing circuit further including at least:

a communication interface (ΓΝΤ) with the inverters, and

a memory (STO) for recording the measurements of maximum powers produced successively by the inverters.